Subband domain echo masking for improved duplexity of spectral domain echo suppressors

Description

Technical Field

[0001] The present document relates to echo cancellation and/or echo suppression. In particular, the present document relates to a method and system for improving the perceived duplexity of handsfree telephone applications.

Background

[0002] Electronic devices such as cordless and/or wireless telephones are often used in a handsfree mode where an audio signal received from the far-end (referred to as the receive signal or the far-end signal) is rendered by a loudspeaker of the electronic device. An echo of the rendered far-end signal may be captured together with an audio signal from the near-end (referred to as the near-end signal) by a microphone of the electronic device, thereby yielding the transmit signal which is to be transmitted to the sender of the far-end signal. As such, the transmit signal may comprise an echo of the far-end signal which may lead to annoying artefacts for the sender of the far-end signal.

[0003] Typical electronic devices comprise echo suppression units or echo suppression circuits for removing the echo from the transmit signal. These echo suppression circuits typically attenuate the transmit signal depending on the received far-end signal. As a result, the transmit signal may be strongly attenuated in the case of a strong far-end signal (e.g. when the far-end signal is comprised of high speech activity), thereby leading to a reduction of duplexity (i.e. a reduction of the capability of an undisturbed two-way (listening and talking) voice communication).

[0004] WO2012/166092A1 describes a method for controlling the suppression gain in acoustic echo control. US2012/0237047A1 describes nonlinear reference signal processing for echo suppression. EP1855456A1 describes a method for echo reduction in time-variant systems.

[0005] The present document addresses the above mentioned short-coming of echo suppression circuits and describes a method and a system for increasing duplexity, while at the same time keeping the artefacts caused by the echo low.

Summary

[0006] The invention is set out in the appended set of claims. The embodiments and/or examples of the following description which are not covered by the appended claims are considered as not being part of the present invention.

[0007] According to an aspect, an echo suppression circuit for an electronic device (e.g. comprising a cordless telephone function or a wireless telephone function) is described. The electronic device comprises a loudspeaker for rendering a receive signal received from a corresponding device at the far-end of a communication network. The receive signal may comprise a voice (or speech) signal originating from a user of the corresponding device at the far-end of the communication network. Furthermore, the electronic device comprises a microphone for capturing the transmit signal which is to be transmitted to the corresponding device at the far-end of the communication network. The transmit signal typically comprises an echo of the rendered receive signal (also referred to as the rendered far-end signal). Furthermore, the transmit signal may comprise a near-end signal, wherein the near-end signal typically originates from a user of the electronic device. By way of example, the near-end signal may comprise a voice (or speech) signal originating from the user of the electronic device.

[0008] The echo suppression circuit is configured to attenuate one or more subbands (e.g. spectral subbands) of the transmit signal (or subbands of a signal derived from the transmit signal). The amount of attenuation may be proportional to the content of far-end signal (e.g. echo) in the particular subband, and typically differs per subband. In particular, the echo suppression circuit is configured to determine a subband masking weight, wherein the subband masking weight is indicative of a masking of the echo of the far-end signal by the near-end signal in the subband. The attenuation applied by the echo suppression circuit onto the subband of the transmit signal may depend on the subband masking weight. In particular, the subband masking weight may be determined such that the attenuation applied to the subband of the transmit signal increases, if the masking of the echo of the far-end signal by the near-end signal in the subband decreases and/or vice versa. As such, the subband masking weight may influence the degree of attenuation applied by the echo suppression circuit, as a function of the masking of the echo in the transmit signal by the near-end signal.

[0009] For determining the subband masking weight, the echo suppression circuit may be configured to determine a subband far-end indicator of a voice (or speech) activity in the far-end signal. The subband far-end indicator may be determined based on the far-end signal in the subband. In an embodiment, the subband far-end indicator is determined solely based on the far-end signal in the subband. By way of example, the subband far-end indicator may correspond to or may comprise an estimate of a level of a voice signal comprised within the far-end signal. Alternatively or in addition, the subband far-end indicator may correspond to or may comprise a speech likelihood of the far-end signal (i.e. an indication of the likelihood that the far-end signal comprises a speech signal).

[0010] In a similar manner, the echo suppression circuit may be configured to determine a subband near-end indicator of a voice (or speech) activity of the near-end signal. The subband near-end indicator may be determined based on the far-end signal and based on the transmit signal in the subband. In an embodiment, the subband near-end indicator is determined solely based on the far-end signal and the transmit signal in the subband. By way of example, the subband near-end indicator may correspond to or may comprise an estimate of a level of a voice signal comprised within the near-end signal. Alternatively or in addition, the subband near-end indicator may correspond to or may comprise a speech likelihood of the near-end signal (i.e. an indication of the likelihood that the near-end signal comprises a speech signal).

[0011] The echo suppression circuit may then be configured to determine the subband masking weight, based on the subband far-end indicator and based on the subband near-end indicator. In particular, the echo suppression circuit may be configured to determine the subband masking weight based on a ratio of the subband far-end indicator and the subband near-end indicator. Furthermore, the subband masking weight may be determined based on a correction factor adapted to tune an influence of the masking of the echo of the far-end signal by the near-end signal on the subband attenuation.

[0012] In addition, the echo suppression circuit may be configured to determine a subband attenuation for the transmit signal in the subband. The subband attenuation may be determined based on the subband masking weight. As such, the subband attenuation may depend on the masking of the echo in the transmit signal by the near-end signal. Furthermore, the subband attenuation may be determined based on an estimate of a level of the echo of the far-end signal in the subband. The estimate of the level of the echo of the far-end signal in the subband may correspond to a level of the far-end signal in the subband multiplied by the subband echo path filter coefficient, wherein the subband echo path filter coefficient may be determined based on a correlation between the transmit signal and the far-end signal in the subband.

[0013] Subsequent to determining the subband attenuation, the echo suppression circuit may be configured to attenuate the subband of the transmit signal using the determined subband attenuation. It should be noted that the determined subband attenuation may be time-variant and/or frequency-variant. As such, different subbands of the transmit signal may be attenuated differently. In a similar manner, different temporal excerpts (e.g. different samples) of the subbands of the transmit signal may be attenuated differently.

[0014] As indicated above, the echo suppression circuit may be configured to operate on a plurality of subbands of the transmit signal. In particular, the echo suppression circuit may be configured to determine a plurality of subband masking weights and configured to determine a plurality of subband attenuations for attenuating the plurality of different subbands of the transmit signal. In an embodiment, the echo suppression circuit is configured to determine a plurality of subband far-end indicators based on a corresponding plurality of subbands of the far-end signal. In a similar manner, the echo suppression circuit may be configured to determine a plurality of subband near-end indicators based on a corresponding plurality of subbands of the far-end signal and based on a corresponding plurality of subbands of the transmit signal. Using this information, the echo suppression circuit may determine the plurality of subband masking weights, based on the plurality of subband far-end indicators and based on the plurality of subband near-end indicators. Furthermore, the echo suppression circuit may determine the plurality of subband attenuations for the transmit signal in the plurality of subbands, based on a respective plurality of estimates of the level of the echo of the respective plurality of subbands of the far-end signal and/or based on the respective plurality of subband masking weights. In addition, the echo suppression circuit may attenuate the plurality of subbands of the transmit signal using the respective plurality of determined subband attenuations, thereby attenuating the different subbands differently.

[0015] The echo suppression circuit may comprise a Time-to-Frequency Transform unit configured to determine a plurality of frequency components of the far-end signal and of the transmit signal, respectively. For this purpose, the Time-to-Frequency Transform unit may make use of a Time-to-Frequency Transform such as a Fast Fourier Transform. Furthermore, the echo suppression unit may comprise a Frequency-to-Subband unit configured to determine the reduced plurality of subbands of the far-end signal and of the transmit signal from the plurality of frequency components of the far-end signal and of the transmit signal, respectively. By way of example a plurality of frequency components may be combined to a subband, in order to reduce the computational complexity of the echo suppression circuit. The echo suppression circuit may be configured to determine a plurality of attenuations for the plurality of frequency components of the transmit signal by interpolating the plurality of subband attenuations. In other words, the individual attenuations for the individual frequency components may be determined by interpolating the subband attenuations which are determined for the respective subbands. The interpolation leads to a smoothing of the applied attenuation, thereby avoiding artifacts which might be caused by applying different subband attenuations to the different subbands of the transmit signal. The echo suppression circuit may further comprise an inverse Time-to-Frequency Transform unit configured to determine an echo suppressed transmit signal from the plurality of attenuated subbands of the transmit signal. In particular, the echo suppressed transmit signal may be determined from the plurality of attenuated frequency components of the transmit signal.

[0016] As indicated above, the echo suppression circuit may be configured to determine a subband echo path filter coefficient based on a correlation between the transmit signal and the far-end signal in the subband. The subband echo path filter coefficient may be indicative of the echo path from the loudspeaker of the electronic device to the microphone of the electronic device. In particular, the subband echo path filter coefficient may be indicative of how much energy of the rendered far-end signal is coupled into the transmit signal as an echo. In such cases, the estimate of the level of the echo of the far-end signal in the subband may correspond to a level of the far-end signal in the subband multiplied by the subband echo path filter coefficient. Furthermore, the echo suppression circuit may be configured to determine the subband attenuation by weighting the subband echo gain filter with the subband masking weight.

[0017] It should be noted that "levels" of signals in the present document may be determined based on (averaged) squared samples of the signal or based on a subband energy.

[0018] According to another aspect, an electronic device (e.g. comprising a telephone application) is described. The electronic device comprises a loudspeaker adapted to render a far-end signal received at the electronic device. Furthermore, the electronic device comprises a microphone adapted to capture a transmit signal, wherein the transmit signal may comprise an echo of the rendered far-end signal and a near-end signal. In addition, the electronic device comprises an echo suppression circuit adapted to attenuate a subband of a signal derived from the transmit signal, thereby yielding an echo suppressed transmit signal. The echo suppression circuit may be configured as described in the present document. A transmission unit comprised within the electronic device may be used to transmit the echo suppressed transmit signal to a sender of the far-end signal.

[0019] The electronic device may further comprise an echo cancellation unit adapted to determine an estimate of the echo of the rendered far-end signal, based on the received far-end signal. In addition, the echo cancellation unit may be adapted to subtract the estimate of the echo of the rendered far-end signal from the transmit signal, thereby yielding the signal derived from the transmit signal.

[0020] According to another aspect, a method for attenuating a subband of a transmit signal is described. The method comprises capturing the transmit signal using a microphone, wherein the transmit signal comprises an echo of a far-end signal rendered by a loudspeaker, and a near-end signal. The method may comprise determining a subband far-end indicator of a voice activity in the far-end signal based on the far-end signal in the subband. Furthermore, the method may comprise determining a subband near-end indicator of a voice activity of the near-end signal, based on the far-end signal and based on the transmit signal in the subband. The method may then proceed in determining a subband masking weight, based on the subband far-end indicator and based on the subband near-end indicator. The subband masking weight may be indicative of a masking in the subband of the echo of the far-end signal by the near-end signal. In addition, the method may comprise determining a subband attenuation for the transmit signal in the subband based on an estimate of a level of the echo of the far-end signal in the subband and based on the subband masking weight. The method may then proceed with attenuating the subband of the transmit signal using the determined subband attenuation.

[0021] According to a further aspect, a software program is described. The software program may be adapted for execution on a processor and for performing the method steps outlined in the present document when carried out on the processor.

[0022] According to another aspect, a storage medium is described. The storage medium may comprise a software program adapted for execution on a processor and for performing the method steps outlined in the present document when carried out on the processor.

[0023] According to a further aspect, a computer program product is described. The computer program may comprise executable instructions for performing the method steps outlined in the present document when executed on a computer.

[0024] It should be noted that the methods and systems including its preferred embodiments as outlined in the present document may be used stand-alone or in combination with the other methods and systems disclosed in this document. Furthermore, all aspects of the methods and systems outlined in the present document may be arbitrarily combined. In particular, the features of the claims may be combined with one another in an arbitrary manner.

Short description of the Figures

[0025] The invention is explained below in an exemplary manner with reference to the accompanying drawings, wherein

Fig. 1 illustrates an example electronic device;

Fig. 2a shows example signal processing for reducing the echo on a transmit path of an electronic device in handsfree mode;

Fig. 2b illustrates a flow chart of an example method for suppressing an echo on the transmit path of an electronic device;

Figs. 3a and 3b illustrate a scheme for increasing the duplexity of an electronic device in handsfree mode.

Detailed Description

[0026] Fig. 1 illustrates an example device 100 with which (and in particular within which) the systems and methods described in the present document may be used. The example device 100 comprises a transducer 107 (in the present example a loudspeaker) which is surrounded by a cabinet (also referred to as an enclosure, casing or housing). The transducer 107 may be used to render a receive signal (e.g. a signal received from a remote calling party) when the device 100 is operated in a handsfree mode. The receive signal may also be referred to as the far-end signal. The example electronic device 100 may be a cordless handset or a wireless telephone. The device 100 may further comprise an earpiece 101 which may be the transducer for rendering the receive signal directly to the ear of a user of the device 100.

[0027] In addition, the device 100 typically comprises a microphone 105 for generating a transmit signal (e.g. to be transmitted from the device 100 to the remote calling party). When operated in the handsfree mode, the microphone 105 may not only capture an acoustic signal generated at the electronic device 100 (e.g. by the user), the microphone 105 may also capture an echo of the far-end signal rendered by the transducer 107. As such, the transmit signal may comprise a near-end signal which is generated at the electronic device 100 (e.g. by the user) and an echo of the far-end signal which is due to an echo path from the transducer 107 to the microphone 105 of the device 100.

[0028] Further components of the electronic device 100 may be a display 102, one or more keys of a keyboard 103, a PCB (Printed Circuit Board) 104, PCB retaining clips 109, Battery clips 111, a battery cover 110, a battery, leads and/or wires 108, and antennas 106. Overall, the electronic device 100 may be subject to various mechanical and acoustic resonances / distortions when rendering the receive signal. The resonances / distortions may be caused by the different components of the device 100, and/or by acoustic chambers formed by the different components.

[0029] Fig. 2a illustrates example signal processing 200 which may be performed within the example device 100 of Fig. 1. Furthermore, an example spectrum 210 of a receive audio signal 320 comprising a single tone (or single sinusoid) at frequency f₀ is illustrated. The receive signal 220 (also referred to as the far-end signal 220) is received at the electronic device 100 (e.g. via the antenna 106). The far-end signal 220 may be submitted to signal processing (not illustrated) on the receive path between the antenna 106 and the transducer 107, e.g. Automatic Gain Control (AGC) for limiting a level of the rendered far-end signal 220, volume control for adjusting a level of the rendered far-end signal 220 and/or equalization for modifying a spectral envelope of the rendered far-end signal 220.

[0030] Subsequent to the optional processing on the receive path, the far-end audio signal 220 is rendered by the transducer 107 of the electronic device 100 which results in a frequency spectrum 211 of the rendered far-end audio signal 221. The rendered far-end signal 221 may be distorted (due to resonances caused at the electronic device 100). This is illustrated by the harmonic frequencies of frequency f₀ comprised within the frequency spectrum 211.

[0031] The rendered far-end signal 221 may be captured by the internal microphone 105 of the electronic device 100 as an echo. Furthermore, the microphone 105 may capture a near-end signal 222. In the illustrated example, the spectrum 212 of the near-end signal 222 comprises a single tone (or single sinusoid) at frequency f₁. As such, the overall transmit signal 223 captured by the microphone 105 comprises a captured echo of the rendered far-end signal 221 and the captured version of the near-end signal 222. The spectrum 213 of the transmit signal 223 comprises frequency components of the near-end signal 222 (at frequency f₁), frequency components of the far-end signal 220 (at frequency f₀).

[0032] The processing 200 may comprise an echo cancellation (e.g. within an echo cancellation unit 201) which is configured to estimate the echo of the far-end signal 210 comprised within the transmit signal 223. Example echo cancellation units 201 make use of one or more of the following estimation algorithms: LMS (Least Mean Square), nLMS (normalized LMS), IPnLMS (Improved Proportionate normalized Least-Mean-Square), APA (Affine Projection Algorithm), RLS (Recursive Least Square). The estimation algorithm may be used to estimate the echo of the far-end signal 210 based on the far-end signal 210. The echo cancellation unit 201 is typically configured to (partly) cancel the frequency components of the far-end signal 220 (at frequency f₀) comprised within the transmit signal 223 by subtracting the echo estimate from the transmit signal 223, thereby yielding an echo compensated transmit signal 223. The spectrum 214 of the echo compensated transmit signal 223 is shown in Fig. 2a. It can be seen that the energy of the frequency component at f₀ is reduced as a result of the echo compensation performed in unit 201.

[0033] However, the echo cancellation unit 201 is typically only able to (partly) cancel the fundamental frequency f₀ of the distorted spectrum 211 of the rendered far-end signal 221, since the fundamental frequency f₀ is the only frequency present in the 'reference' input 220 (i.e. in the receive audio signal) of the echo cancellation unit 201. In other words, the echo cancellation unit 201 is typically only able to cancel the frequency components already comprised within the far-end signal 220. Hence, after cancelling the ground harmonic at frequency f₀ (i.e. after cancelling the frequency components comprised within the far-end signal 220), the output signal of the echo cancellation unit 201 typically comprises the distortion components of the echo signal (illustrated by the echo cancelled spectrum 214) as well as the frequency components of the near-end signal 222. Furthermore, the output signal of the echo cancellation unit 201 may still comprise a residual of the frequency components comprised within the far-end signal 220.

[0034] The signal processing may comprise a spectral-domain echo suppression unit 202. Typically, the spectral-domain echo suppression unit 202 considers the energy in the 'reference' signal 220 (i.e. in the far-end audio signal) after dividing the 'reference' signal 220 into L spectral subbands (L being an integer greater than one). For each subband the suppressor 202 estimates the echo energy and subtracts the estimated echo energy within the subband from the transmit signal 223 (e.g. from the transmit signal 223 subsequent to the echo cancellation unit 201). This is typically achieved by the application of an attenuation factor proportional to the amount of echo content in each subband. When a subband contains both echo and near-end speech it is attenuated due to the echo, which leads to reduced duplexity.

[0035] In more detail, the spectral-domain echo suppression unit 202 (also referred to as the Spectral Subtraction Echo Suppressor 202) may make use of an algorithm which predicts the echo energy from the available spectral and temporal properties of the far-end signal 220. The prediction of the echo energy is typically based on an (adaptive) estimation of the coupling between the speaker 107 and the microphone 105. The spectral-domain echo suppression unit 202 subtracts the estimated echo energy from the transmit signal 223, thereby yielding an echo suppressed transmit signal 223.

[0036] An example method 250 for echo suppression performed within the spectral-domain echo suppression unit 202 is illustrated in Fig. 2b. In step 251, the transmit signal t(n) 223 (or an echo compensated version thereof) and the far-end signal r(n) 220 are converted from the time-domain into the frequency-domain using a Time-to-Frequency Transform (e.g. a Fast Fourier Transform), thereby yielding T(k,n) and R(k,n). In the above notation, n is a time or sample index and k = 1,..., K is a frequency bin index (e.g. K = 128). Subsequently, some of the frequency bins k may be combined, in order to form a reduced number L (e.g. L = 12) of subbands R(l,n) and T(l,n), with l = 1,..., L being the subband index (step 252). The combination of a plurality of frequency bins k into a subband may be performed in order to reduce the computational complexity of the echo suppressor 202. Aggregate power levels may be determined for each subband based on the sum of a plurality of frequency bins comprising the subband, e.g. R(l,n) = ∑R(k,n) and T(l,n) = ∑T(k,n).

[0037] In method step 253 the echo path properties are estimated for each of the subbands l = 1,..., L. The echo path properties may describe a relationship between a subband R(l,n) of the far-end signal 220 and a corresponding subband T(l,n) of the transmit signal 223 (i.e. of the input signal to the echo suppressor 202) in the absence of a near-end signal 222. The echo path property of a subband l = 1,...,L may be described e.g. by a filter H(l,n) such as a Finite Impulse Response (FIR) filter

with h_m(l,n) being the coefficients of the filter. It should be noted that a time-domain subband canceller may serve as an alternative implementation. In an embodiment, the echo path property of a subband l = 1,..., L is described by a single gain coefficient h₀(l,n), i.e. by an FIR filter with M = 1 coefficients. In another example, M is greater than one. The echo path property of a subband l = 1,..., L may be determined by considering the subband T(l,n) of the transmit signal 223 in the absence of a near-end signal 222 in conjunction with the corresponding subband R(l,n) of the far-end signal 220. A cross-correlation between R(l,n) and T(l,n), e.g. C(l,0) = ∑R(l,n)·T(l,n) across an observation window of a plurality of samples, may then provide an indication of the gain coefficient h₀(l,n). In other words, the echo path properties, e.g. described by the echo path filter coefficient h₀(l,n), may be determined based on the cross-correlation between R(l,n) and T(l,n).

[0038] The echo path properties of the subbands l = 1,..., L may be used to determine an estimate of the echo energy within the respective subbands. In an embodiment, the echo path properties are described by a filter H(l,n) and an estimate of the echo energy E(l,n) comprised within a subband T(l,n) of the transmit signal 223 may be determined by filtering the subband R(l,n) of the far-end signal 220 using the filter H(l,n) describing the echo path properties. The estimated echo energy E(l,n) within a subband l may then be substracted from the subband T(l,n) of the transmit signal 223, and divided by T(l,n) to yield a subband echo gain filter, G(l,n). This subband echo gain filter G(l,n) can subsequently be applied during the final subband-to-frequency conversion, resulting in an echo suppressed transmit signal, to which an inverse frequency-to-time transformation (e.g. inverse Fourier Transform) can be applied. 223. The echo energy compensation is performed in method step 256.

[0039] It should be noted that the use of different echo path properties, e.g. the use of different filters H(l,n) or different gain coefficients h₀(l,n), for different subbands may lead to discontinuities at the subband edges for the subband echo gain filter G(l,n). In view of this, it may be beneficial to smoothen the echo path properties, e.g. the filters H(l,n) or echo path filter coefficients h₀(l,n). The smoothening may be performed in an optional method step 255. In an embodiment, the smoothening comprises the determination of different echo path properties for the different frequency bins k = 1,..., K from the echo path properties for the different subbands l = 1,..., L. The echo path properties for the different frequency bins k = 1,..., K may be determined by interpolating the echo path properties for the different subbands l = 1,..., L, such that smooth transitions are obtained from one frequency bin to an adjacent frequency bin. By way of example, the echo path filter coefficients h₀(l,n) for l = 1,...,L may be interpolated to determine gain coefficients h₀(k,n) for k = 1,...,K.

[0040] The smoothened (e.g. interpolated) echo path properties (e.g. the echo path filter coefficients h₀(k,n)) may be applied to the subbands R(l,n) (or the frequency bins R(k,n)) of the far-end signal 220, thereby yielding an estimate of the echo energy E(l,n) within a subband l (or an estimate of the echo energy E(k,n) within a frequency bin k). The estimated echo energy E(l,n) within a subband l may then be subtracted from the subband T(l,n) of the transmit signal 223, and the difference may be divided by the subband T(l,n) of the transmit signal 223 to yield a subband echo gain filter G(l,n). As indicated above, the echo energy compensation is performed in method step 256. The subband echo gain filter G(l,n) may be applied to the subband T(l,n) of the transmit signal, thereby yielding an echo suppressed subband T̂(l,n) of the transmit signal. Using a subband to bin transform, the echo suppressed frequency bins T̂(k,n) may be determined (method step 257).

[0041] The echo suppressed frequency bins T̂(k,n) of the transmit signal 223 are converted back into the time domain (method step 258) using an inverse Time-to-Frequency Transform (e.g. an inverse Fast Fourier Transform). As a result, an echo suppressed version of the transmit signal 223 is obtained at the output of the spectral-domain echo suppression unit 202. It should be noted that the processing 200 may comprise further post processing (e.g. unit 203) and the processed transmit signal 223 may be encoded for transmission over a telecommunication network.

[0042] The echo suppression applied within the spectral-domain echo suppression unit 202 is typically such that all frequency components of the transmit signal 223 (even the frequency components which are due to the near-end signal 222) are attenuated. This leads to a reduction of the full duplexity of the electronic device 100, because the components of the near-end signal 222 within the transmit signal 223 may be attenuated such that the near-end signal 222 is not (or not sufficiently) audible by a user at the far-end. In view of the above, it is beneficial to modify the processing of the spectral-domain echo suppression unit 202 in order to improve the full duplexity of the electronic device 100 (in particular, when operated in handsfree mode).

[0043] In the present document, a method and system is described, wherein the method and system automatically reduce echo attenuation applied to the transmit path of the electronic device 100, whenever the echo energy is masked by the presence of a near-end signal 222. In other words, it is proposed to consider masking by the near-end signal 222 of an echo comprised within the transmit signal 223. If it is determined that an echo comprised within the transmit signal 223 is masked by the near-end signal 222, the echo attenuation applied within the spectral-domain echo suppression unit 202 may be reduced, thereby increasing duplexity. At the same time, the audible echo perceived within the transmit signal 223 remains small, due to the masking of the echo by the near-end signal 222.

[0044] The masking effect may be taken into account when performing echo suppression using an additional masking weight which is applied when determining the estimated echo E(l,n) within a subband l and at the time or sample index n. The masking weight may be determined in the context of step 254 of method 250 of Fig. 2b.

[0045] In an embodiment, a masking weight w(l,n) is determined for each of the subbands l = 1,...,L. The masking weight w(l,n) may be a function of (e.g. a function of the ratio of) an indicator a_NE(l,n) of the audio activity (e.g. a speech activity) of the near-end signal 222 and an indicator a_FE(l,n) of the audio activity (e.g. a speech activity) of the far-end signal 222, i.e. w(l,n) = f(a_NE(l,n),a_FE(l,n)), wherein the operator f() represents an arbitrary function. Furthermore, the masking weight w(l,n) may be a function of a correction factor α which can be used to emphasize or limit the effect of the masking weight on the echo suppression. In other words, the correction factor α can be used to emphasize or limit the masking weight, thereby emphasizing or limiting the effect of the masking weight on the echo suppression.

[0046] As indicated above, the masking weight w(l,n) may be used to take into account masking effects, when determining the echo suppressed near-end signal. This may be done, e.g. by using the following equation for determining the subband echo gain filter:

[0047] The masking weight w(l,n) may be defined such that the masking weight w(l,n) increases with an increasing masking effect. In particular, the masking weight w(l,n) may take on a maximum value when the transmit signal fully masks the echo, and a minimum value when there is no masking. A masking weight w(l,n) less than 1 may be applied to the subband gain filter coefficients, G(l,n) (as indicated by the above mentioned equation), in order to reduce the estimated echo energy E(l,n) within a subband l. In particular, w(l,n) is applied to the echo gain filter coefficients, such that as the masking affect increases, w(l,n) increases and vice versa, as can be seen by the above mentioned equation for G(l,n).

[0048] It should be noted that other definitions of the masking weight w(l,n) are possible. By way of example, the subband echo gain filter may be determined using an alternative equation:

In such a case, the masking weight w(l,n) may be defined such that the masking weight w(l,n) decreases with an increasing masking effect. In general terms, the masking weight w(l,n) is configured to decrease the amount of attenuation of a subband l = 1,..., L, of the transmit signal 223, if it is determined that the near-end signal 222 masks an echo of the far-end signal 220 comprised within the transmit signal 223. The amount of attenuation is decreased, as the masking effect increases.

[0049] The effect of applying masking weights when determining the estimated echo energy (and the corresponding echo suppression) can be illustrated using the spectra 214, 215 of the transmit signal 223 at the input and at the output of the echo suppression unit 202, respectively. The near-end signal 222 only comprises a frequency component at the frequency f₁, whereas the far-end signal 220 only comprises a frequency component at the frequency f₀. The frequencies f₀ and f₁ may be different from one another. The near-end signal 222 masks any echo components in the subband l₁ comprising the frequency f₁. In other words, the near-end signal 222 masks echo components in a frequency range which is close to the frequency component at the frequency f₁ of the near-end signal 222. On the other hand, no masking occurs in other subbands. As such, the masking weight w(l₁,n) of subband l₁ would be higher than the masking weight w(l,n) for the other subbands. This results in reduced attenuation for the subband l₁ comprising the frequency f₁ compared to the other subbands, thereby leading to a reduced attenuation for the near-end signal 222, i.e. thereby increasing the duplexity of the electronic device 100. This is illustrated by the spectrum 215 of the echo suppressed transmit signal 223, where the frequency component at frequency f₁ remains substantially un-attenuated, whereas the other frequency components (caused by the echo) are attenuated.

[0050] In other words, the far-end signal 220 may generate non-linear harmonies within the echo signal. The near-end signal 222 comprising a frequency component at the frequency f₁ will mask harmonics within the echo signal that are in the same frequency band as frequency f₁. The methods described in the present document help relax the regular filters, the non-linear post filter and the mechanical post filter.

[0051] As indicated above, the masking weight w(l,n) may be a function of an appropriate indicator a_NE(l,n) of the audio activity of the near-end signal 222 and an appropriate indicator a_FE(l,n) of the audio activity of the far-end signal 220, and possibly a correction factor α. The indicators a_NE(l,n) and/or a_FE(l,n) may be determined based on an estimate of the energy level of an audio signal (e.g. a voice signal) comprised within the near-end signal (222) and/or an estimate of the energy level of an audio signal (e.g. a voice signal) comprised within the far-end signal (223). Alternatively or in addition, the indicators a_NE(l,n) and/or a_FE(l,n) may be determined based on an estimate of the speech likelihood of the near-end signal (222) and/or an estimate of the speech likelihood of the far-end signal (223). Various schemes for determining the energy level and/or the speech likelihood may be used. In the following an example scheme is described.

[0052] In an embodiment, the indicator a_FE(l,n) is determined based on the far-end signal 220 on the receive path. A level of the background (stationary) noise comprised within the far-end signal 220 may be estimated e.g. using an averaging filter across a plurality of samples of the far-end signal 220. The estimated level of the background noise may be subtracted from the level of the far-end signal 220 on the receive path, thereby yielding an estimate of the level of a speech signal comprised within the far-end signal 220. It should be noted that the above mentioned operations may be performed for each subband l = 1,..., L, separately, thereby yielding L estimates of the level of a speech signal comprised within the L subbands of the far-end signal 220, respectively. The indicator a_FE(l,n) may be determined based on (e.g. may be equal to) the estimate of the level of a speech signal comprised within the subband l of the far-end signal 220.

[0053] The indicator a_NE(l,n) may determined based on the transmit signal 223 on the transmit path (e.g. based on the echo compensated transmit signal 223 at the input of the echo suppression unit 202). A level of the background (stationary) noise comprised within the transmit signal 223 may be estimated e.g. using an averaging filter across a plurality of samples of the transmit signal 223. The estimated level of the background noise may be subtracted from the level of the transmit signal 223 on the transmit path, thereby yielding an estimate of the level of a noise compensated transmit signal 223. Furthermore, an estimate of the level of the echo (due to the far-end signal 220) comprised within the transmit signal 223 may be determined based on a cross-correlation of the far-end signal 220 and the transmit signal 223, e.g. based on C(l,0) = ∑R(l,n)·T(l,n) across an observation window of a plurality of samples. The cross-correlation may then be subtracted from the noise compensated transmit signal 223 to yield an estimate of the level of a speech signal comprised within the near-end signal 222. It should be noted that the above mentioned operations may be performed for each subband l = 1,..., L, separately, thereby yielding L estimates of the level of a speech signal comprised within the L subbands of the near-end signal 222, respectively. The indicator a_NE(l,n) may be determined based on (e.g. may be equal to) the estimate of the level of a speech signal comprised within the subband l of the near-end signal 222.

[0054] As indicated above, the masking weight may be determined as w(l,n) = f(a_NE(l,n),a_FE(l,n),α). Example functions for determining the masking weight are:

[0055] In the first example w(l,n) is continuous and proportional to the near-end masking affect, as indicated by a(l,n).

[0056] In the second example, the near-end indicator is compared to a programmable near end threshold, t_NE(l,n)). When the near-end indicator exceeds the threshold t_NE(l,n)), weight w(l,n) increases, thereby reducing the echo attenuation factor G(l,n). Otherwise, the value of w(l,n) is unity, resulting in no reduction to gain filter G(l,n).

[0057] It should be noted that the masking weight may de determined using other functions, depending on how the masking weight is used to modify the degree of suppression (i.e. the degree of attenuation). By way of example, the masking weight may be determined as

such that a low masking weight indicates a strong masking effect and vice versa. As indicated above, the indicators a_NE(l,n) and/or a_FE(l,n) may be determined based on a speech likelihood and/or based on a level of the near-end signal 222 and the far-end signal 220, respectively.

[0058] The masking weight w(l,n) may be applied to the subband echo gain filter coefficients, yielding modified gain filter coefficients

for l = 1,..., L. As a result, the echo suppression unit 202 attenuates less energy when there is near-end speech that may be masking the actual echo energy. As indicated above, the implementation of the masking weight w(l,n) may rely on a ratio between the speech activity of the near-end and far-end signals. Consequently, the effect of a reduced attenuation of energy may be stronger whenever the near-end speech activity is stronger than the far-end speech activity. And inversely, the weighing may be reduced automatically, if and when there is not enough near-end speech activity to mask the echo energy from the far-end signal..

[0059] It should be noted that gain smoothing (step 255) may be applied to the modified gain filter coefficients G(l,n) in a similar manner as described above (e.g. by interpolating the modified gain coefficients G(l,n)), in order to yield modified gain coefficients for each frequency bin k = 1,..., K.

[0060] The masking effect and the reduced echo suppression are illustrated in Figs. 3a and 3b. Fig. 3a shows the indicator a_FE(l,n) 301 of the audio activity of an example far-end signal 220. In the illustrated example, the indicator a_FE(l,n) 301 corresponds to a level of the speech or to a speech likelihood of the far-end signal 220. The indicator a_FE(l,n) 301 of the audio activity of the example far-end signal 220 varies along the time line (i.e. along the sample index n).

[0061] There may be a (programmable) limit to the strength of the masking weight. By way of example, the masking weight may be such that it may not relax attenuation more than a pre-determined maximum amount.

[0062] Fig. 3b illustrates the indicator a_NE(l,n) 312 of the audio activity of an example near-end signal 222. In the illustrated example, the indicator a_NE(l,n) 312 corresponds to a level of the speech or to a speech likelihood of the near-end signal 222. The indicator a_NE(l,n) 312 of the audio activity of the example near-end signal 222 varies along the time line (i.e. along the sample index n). Furthermore, Fig. 3b shows a corresponding indicator 311 of the audio activity of the echo of the far-end signal 220 comprised within the transmit signal 223. The corresponding indicator 311 is typically correlated with the indicator a_FE(l,n) 301 of the audio activity of the example far-end signal 220 shown in Fig. 3a. It can be seen that within the time interval 313, the audio activity of the near-end signal 222 masks the audio activity of the echo in the transmit signal 223. This masking effect is reflected by the indicators a_FE(l,n) 301 and a_NE(l,n) 312, leading to a reduced echo suppression in the time interval 313, as illustrated by the curve 324 of the non-weighted echo gain filter coefficient and the curve 325 of the weighted echo gain filter coefficient G(l,n).

[0063] The weighted echo gain filter coefficient G(l,n) leads to reduced echo suppression, i.e. to a reduced attenuation of the transmit signal 223 during the time interval 313, thereby increasing the duplexity of the electronic device 100.

[0064] In the present document, a method and system for improving the duplexity of an electronic device operated in handsfree mode have been described. The described method and system exploit the masking of an echo comprised within a transmit signal by a near-end signal comprised within the transmit signal. For this purpose, a masking weight is determined which reduces the attenuation of a spectral domain echo suppressor (or a subband domain or time domain implementation) as a function of the degree of the masking effect. By reducing the attenuation in spectral bands with masking and/or during time periods with masking, the overall degree of attenuation can be reduced, without increasing the perceived artefacts caused by an echo comprised within the transmit signal. As a result, the duplexity of the electronic device is increased.

[0065] It should be noted that the description and drawings merely illustrate the principles of the proposed methods and systems. Furthermore, all examples and embodiment outlined in the present document are principally intended expressly to be only for explanatory purposes to help the reader in understanding the principles of the proposed methods and systems.

Claims

1. An echo suppression circuit (202) for an electronic device (100) comprising a loudspeaker (107) and a microphone (105); wherein the echo suppressing circuit (202) is configured to attenuate a subband of a transmit signal (223); wherein the transmit signal (223) is either captured by the microphone (105) or derived from a signal captured by the microphone (105); wherein the transmit signal (313) comprises an echo of a far-end signal (220) rendered by the loudspeaker (107) and a near-end signal (222); characterized in that the echo suppressing circuit (202) is configured to

- determine a subband far-end indicator of a voice activity in the far-end signal (220) based on the far-end signal (220) in the subband; wherein the subband far-end indicator comprises an estimate of a level of a voice signal comprised within the far-end signal and/or an indication of the likelihood that the far-end signal comprises a speech signal;

- determine a subband near-end indicator of a voice activity of the near-end signal (222) based on the far-end signal (220) in the subband and based on the transmit signal (223) in the subband; wherein the subband near-end indicator comprises an estimate of a level of a voice signal comprised within the near-end signal and/or an indication of the likelihood that the near-end signal comprises a speech signal;

- determine a subband masking weight, based on the subband far-end indicator and based on the subband near-end indicator; wherein the subband masking weight is indicative of a masking of the echo of the far-end signal (220) by the near-end signal (222) in the subband;

- determine a subband echo path filter coefficient based on a correlation between the transmit signal (223) in the subband and the far-end signal (220) in the subband;

- determine an estimate of the level of the echo of the far-end signal (220) in the subband by multiplying a level of the far-end signal (220) in the subband by the subband echo path filter coefficient;

- determine a subband attenuation for the transmit signal (223) in the subband by weighting the subband echo path filter coefficient with the subband masking weight; and

- attenuate the subband of the transmit signal (223) by subtracting the level of the far-end signal (220) in the subband multiplied by the subband attenuation from a level of the transmit signal (223) in the subband.

2. The echo suppression circuit (202) of claim 1, wherein the circuit (202) is configured to

- determine a plurality of subband far-end indicators based on a corresponding plurality of subbands of the far-end signal (220);

- determine a plurality of subband near-end indicators based on a corresponding plurality of subbands of the far-end signal (220) and based on a corresponding plurality of subbands of the transmit signal (223);

- determine a plurality of subband masking weights, based on the plurality of subband far-end indicators and based on the plurality of subband near-end indicators;

determining, for each subband, a subband echo path filter coefficient based on a correlation between the transmit signal in said subband and the far-end signal in said subband;

- determine a plurality of subband attenuations for the transmit signal (223) in the plurality of subbands by weighting, for each subband, the subband echo path filter coefficient of said subband with the subband masking weight of said subband;

- attenuate the plurality of subbands of the transmit signal (223) using the corresponding plurality of determined subband attenuations.

3. The echo suppression circuit (202) of claim 2, wherein the circuit (202) comprises

- a Time-to-Frequency Transform unit configured to determine a plurality of frequency components of the far-end signal (220) and of the transmit signal (223); and

- a Frequency-to-Subband unit configured to determine the plurality of subbands of the far-end signal (220) and of the transmit signal (223) from the plurality of frequency components of the far-end signal (220) and of the transmit signal (223), respectively.

4. The echo suppression circuit (202) of claim 3, wherein the circuit (202) is configured to

- determine a plurality of attenuations for the plurality of frequency components of the transmit signal (223) by interpolating the plurality of subband attenuations.

5. The echo suppression circuit (202) of any of claims 2 to 4, wherein the circuit (202) comprises

- an inverse Time-to-Frequency Transform unit configured to determine an echo suppressed transmit signal (223) from the plurality of attenuated subbands of the transmit signal (223).

6. The echo suppression circuit (202) of any previous claim, further configured to determine the subband masking weight based on a ratio of the subband far-end indicator and the subband near-end indicator.

7. The echo suppression circuit (202) of claim 6, further configured to determine the subband masking weight based on a correction factor adapted to tune an influence of the masking of the echo of the far-end signal (220) by the near-end signal (222) on the subband attenuation.

8. The echo suppression circuit (202) of any previous claim, wherein the determined subband attenuation is time-variant and/or frequency-variant.

9. The echo suppression circuit (202) of any previous claim, further configured to determine the subband masking weight such that the attenuation applied to the subband of the transmit signal (223) increases, if the masking of the echo of the far-end signal (220) by the near-end signal (222) in the subband decreases and vice versa.

10. An electronic device (100) comprising

- a loudspeaker (107) adapted to render a far-end signal (220) received at the electronic device (100);

- a microphone (105) adapted to capture a signal (223); wherein the captured signal (223) comprises an echo of the rendered far-end signal (220) and a near-end signal (222);

- an echo suppression circuit (202) of any of claims 1 to 9, adapted to attenuate a subband of a signal derived from the captured signal (223), thereby yielding an echo suppressed transmit signal; and

- a transmission unit configured to transmit the echo suppressed transmit signal to a sender of the far-end signal (220).

11. The electronic device (100) of claim 10, further comprising an echo cancellation unit (201) adapted to

- determine an estimate of the echo of the rendered far-end signal (220), based on the received far-end signal (220); and

- subtract the estimate of the echo of the rendered far-end signal (220) from the captured signal (223), thereby yielding the signal derived from the captured signal (223).

12. A method for attenuating a subband of a transmit signal (223); the method being performed by an echo suppressing circuit and comprising

- capturing the transmit signal (223) using a microphone (105); wherein the transmit signal (313) comprises an echo of a far-end signal (220) rendered by a loudspeaker (107) and a near-end signal (222); said method being characterized by:

- determining a subband far-end indicator of a voice activity in the far-end signal (220) based on the far-end signal (220) in the subband; wherein the subband far-end indicator comprises an estimate of a level of a voice signal comprised within the far-end signal and/or an indication of the likelihood that the far-end signal comprises a speech signal;

- determining a subband near-end indicator of a voice activity of the near-end signal (222) based on the far-end signal (220) in the subband and based on the transmit signal (223) in the subband; wherein the subband near-end indicator comprises an estimate of a level of a voice signal comprised within the near-end signal and/or an indication of the likelihood that the near-end signal comprises a speech signal;

- determining a subband masking weight, based on the subband far-end indicator and based on the subband near-end indicator; wherein the subband masking weight is indicative of a masking of the echo of the far-end signal (220) by the near-end signal (222) in the subband;

- determining a subband echo path filter coefficient based on a correlation between the transmit signal (223) in the subband and the far-end signal (220) in the subband;

- determining an estimate of the level of the echo of the far-end signal (220) in the subband by multiplying a level of the far-end signal (220) in the subband by the subband echo path filter coefficient;

- determining a subband attenuation for the transmit signal (223) in the subband by weighting the subband echo path filter coefficient with the subband masking weight; and

- attenuating the subband of the transmit signal (223) by subtracting the level of the far-end signal (220) in the subband multiplied by the subband attenuation from a level of the transmit signal (223) in the subband.

Ansprüche

1. Echounterdrückungsschaltkreis (202) für eine elektronische Vorrichtung (100), umfassend einen Lautsprecher (107) und ein Mikrofon (105); wobei der Echounterdrückungsschaltkreis (202) konfiguriert ist, ein Unterband eines Sendesignals (223) zu dämpfen; wobei das Sendesignal (223) entweder durch das Mikrofon (105) erfasst wird oder von einem Signal abgeleitet wird, das durch das Mikrofon (105) erfasst wird; wobei das Sendesignal (313) ein Echo eines Signals (220) am entfernten Ende, das durch den Lautsprecher (107) wiedergegeben wird, und ein Signal (222) am nahen Ende umfasst; dadurch gekennzeichnet, dass der Echounterdrückungsschaltkreis (202) konfiguriert ist zur

- Bestimmung eines Unterbandanzeigers einer Stimmaktivität am entfernten Ende in dem Signal (220) am entfernten Ende auf Grundlage des Signals (220) am entfernten Ende in dem Unterband; wobei der Unterbandanzeiger am entfernten Ende eine Schätzung eines Pegels eines Stimmsignals, das in dem Signal am entfernten Ende enthalten ist, und/oder eine Anzeige der Wahrscheinlichkeit, dass das Signal am entfernten Ende ein Sprachsignal umfasst, umfasst;

- Bestimmung eines Unterbandanzeigers einer Stimmaktivität am nahen Ende des Signals (222) am nahen Ende auf Grundlage des Signals (220) am nahen Ende in dem Unterband und auf Grundlage des Sendesignals (223) in dem Unterband; wobei der Unterbandanzeiger am nahen Ende eine Schätzung eines Pegels eines Stimmsignals, das in dem Signal am nahen Ende enthalten ist, und/oder eine Anzeige der Wahrscheinlichkeit, dass das Signal am nahen Ende ein Sprachsignal umfasst, umfasst;

- Bestimmung eines Unterbandmaskierungsgewichts auf Grundlage des Unterbandanzeigers am entfernten Ende und auf Grundlage des Unterbandanzeigers am nahen Ende; wobei das Unterbandmaskierungsgewicht eine Maskierung des Echos des Signals (220) am entfernten Ende durch das Signal (222) am nahen Ende in dem Unterband anzeigt;

- Bestimmung eines Unterbandechopfadfilterkoeffizienten auf Grundlage einer Korrelation zwischen dem Sendesignal (223) in dem Unterband und dem Signal (220) am entfernten Ende in dem Unterband;

- Bestimmung einer Schätzung des Echopegels des Signals (220) am entfernten Ende in dem Unterband durch Multiplizieren eines Pegels des Signals (220) am entfernten Ende in dem Unterband durch den Unterbandechopfadfilterkoeffizienten;

- Bestimmung einer Unterbanddämpfung für das Sendesignal (223) in dem Unterband durch Gewichtung des Unterbandechopfadfilterkoeffizienten mit dem Unterbandmaskierungsgewicht; und

- Dämpfung des Unterbands des Sendesignals (223) durch Subtrahieren des Pegels des Signals (220) am entfernten Ende in dem Unterband, multipliziert mit der Unterbanddämpfung von einem Pegel des Sendesignals (223) in dem Unterband.

2. Echounterdrückungsschaltkreis (202) nach Anspruch 1, wobei der Schaltkreis (202) konfiguriert ist zur

- Bestimmung mehrerer Unterbandanzeiger am entfernten Ende auf Grundlage entsprechender mehrerer Unterbänder des Signal (220) am entfernten Ende;

- Bestimmung mehrerer Unterbandanzeigers am nahen Ende auf Grundlage entsprechender mehrerer Unterbänder des Signals (220) am entfernten Ende und auf Grundlage entsprechender mehrerer Unterbänder des Sendesignals (223);

- Bestimmung mehrerer Unterbandmaskierungsgewichte auf Grundlage der mehreren Unterbandanzeiger am entfernten Ende und auf Grundlage der mehreren Unterbandanzeiger am nahen Ende;

- Bestimmung eines Unterbandechopfadfilterkoeffizienten für jedes Unterband auf Grundlage einer Korrelation zwischen dem Sendesignal in dem Unterband und dem Signal am entfernten Ende in dem Unterband;

- Bestimmung mehrerer Unterbanddämpfungen für das Sendesignal (223) in den mehreren Unterbändern durch Gewichtung des Unterbandechopfadfilterkoeffizienten des Unterbands für jedes Unterband mit dem Unterbandmaskierungsgewicht des Unterbands;

- Dämpfung der mehreren Unterbänder des Sendesignals (223) unter Verwendung der entsprechenden mehreren der bestimmten Unterbanddämpfungen.

3. Echounterdrückungsschaltkreis (202) nach Anspruch 2, wobei der Schaltkreis (202) umfasst:

- eine Zeit-zu-Frequenz-Transformationseinheit, konfiguriert zur Bestimmung mehrerer Frequenzkomponenten des Signals (220) am entfernten Ende und des Sendesignals (223); und

- eine Frequenz-zu-Unterbandeinheit, konfiguriert zur Bestimmung der mehreren Unterbänder des Signals (220) am entfernten Ende und des Sendesignals (223) von den mehreren Frequenzkomponenten des Signals (220) am entfernten Ende bzw. des Sendesignals (223) .

4. Echounterdrückungsschaltkreis (202) nach Anspruch 3, wobei der Schaltkreis (202) konfiguriert ist zur

- Bestimmung mehrerer Dämpfungen für die mehreren Frequenzbestandteile des Sendesignals (223) durch Interpolieren der Unterbanddämpfungen.

5. Echounterdrückungsschaltkreis (202) nach einem der Ansprüche 2 bis 4, wobei der Schaltkreis (202) umfasst:

- eine invertierte Zeit-zu-Frequenztransformationseinheit, konfiguriert zur Bestimmung eines echounterdrückten Sendesignals (223) von den mehreren gedämpften Unterbändern des Sendesignals (223).

6. Echounterdrückungsschaltkreis (202) nach einem der vorhergehenden Ansprüche, ferner konfiguriert zur Bestimmung des Unterbandmaskierungsgewichts auf Grundlage eines Verhältnisses des Unterbandanzeigers am entfernten Ende und des Unterbandanzeigers am nahen Ende.

7. Echounterdrückungsschaltkreis (202) nach Anspruch 6, ferner konfiguriert zur Bestimmung des Unterbandmaskierungsgewichts auf Grundlage eines Korrekturfaktors, angepasst, um einen Einfluss der Maskierung des Echos des Signals (220) am entfernten Ende durch das Signal (222) am nahen Ende auf die Unterbanddämpfung einzustellen.

8. Echounterdrückungsschaltkreis (202) nach einem der vorhergehenden Ansprüche, wobei die bestimmte Unterbanddämpfung eine Zeitvariante und/oder Frequenzvariante ist.

9. Echounterdrückungsschaltkreis (202) nach einem der vorherigen Ansprüche, ferner konfiguriert zur Bestimmung des Unterbandmaskierungsgewichts, sodass die Dämpfung, die auf das Unterband des Sendsignals (223) angewendet wird, ansteigt, wenn die Maskierung des Echos des Signals (220) am entfernten Ende durch das Signal (222) am nahen Ende in dem Unterband sinkt und umgekehrt.

10. Elektronische Vorrichtung (100), umfassend

- einen Lautsprecher (107), der angepasst ist, ein Signal (220) am entfernten Ende zu rendern, das an der elektronischen Vorrichtung (100) empfangen wird;

- ein Mikrofon (105), das angepasst ist, ein Signal (223) zu erfassen; wobei das erfasste Signal (223) ein Echo eines gerenderten Signals (220) am entfernten Ende und ein Signal (222) am nahen Ende umfasst;

- einen Echounterdrückungsschaltkreis (202) nach einem der Ansprüche 1 bis 9, angepasst zum Dämpfen eines Unterbands eines Signals, das von dem erfassten Signal (223) abgeleitet ist, wodurch ein echounterdrücktes Sendesignal erhalten wird; und

- eine Sendeeinheit, konfiguriert zum Senden des echounterdrückten Sendesignals an einen Sender des Signals (220) am entfernten Ende.

11. Elektronische Vorrichtung (100) nach Anspruch 10, ferner umfassend eine Echounterdrückungseinheit (201), angepasst zur

- Bestimmung einer Schätzung des Echos des gerenderten Signals (220) am entfernten Ende, auf Grundlage des empfangenen Signals (220) am entfernten Ende; und

- Abziehen der Schätzung des Echos des gerenderten Signals (220) am entfernten Ende von dem erfassten Signal (223), wodurch das Signal, das von dem erfassten Signal (223) abgeleitet wird, erhalten wird.

12. Verfahren zum Dämpfen eines Unterbands eines Sendesignals (223); wobei das Verfahren durch einen Echounterdrückungsschaltkreis ausgeführt wird und umfasst:

- Erfassung des Sendesignals (223) unter Verwendung eines Mikrofons (105); wobei das Sendesignal (313) ein Echo eines Signals (220) am entfernten Ende, das durch einen Lautsprecher (107) gerendert wird, und ein Signal (222) am nahen Ende umfasst; wobei dieses Verfahren gekennzeichnet ist durch:

- Bestimmung eines Unterbandanzeigers einer Stimmaktivität am entfernten Ende in dem Signal (220) am entfernten Ende auf Grundlage des Signals (220) am entfernten Ende in dem Unterband; wobei der Unterbandanzeiger am entfernten Ende eine Schätzung eines Pegels eines Stimmsignals, das in dem Signal am entfernten Ende enthalten ist, und/oder eine Anzeige einer Wahrscheinlichkeit, dass das Signal am entfernten Ende ein Sprachsignal umfasst, umfasst;

- Bestimmung eines Unterbandanzeigers einer Stimmaktivität am nahen Ende des Signals (222) am nahen Ende auf Grundlage des Signals (220) am nahen Ende in dem Unterband und auf Grundlage des Sendesignals (223); wobei der Unterbandanzeiger am nahen Ende eine Schätzung eines Pegels eines Stimmsignals, das in dem Signal am nahen Ende enthalten ist, und/oder eine Anzeige einer Wahrscheinlichkeit, dass das Signal am nahen Ende ein Sprachsignal umfasst, umfasst;

- Bestimmung eines Unterbandmaskierungsgewichts auf Grundlage des Unterbandanzeigers am entfernten Ende und auf Grundlage des Unterbandanzeigers am nahen Ende; wobei das Unterbandmaskierungsgewicht eine Maskierung des Echos des Signals (220) am entfernten Ende durch das Signal (222) am nahen Ende in dem Unterband anzeigt;

- Bestimmung eines Unterbandechopfadfilterkoeffizienten auf Grundlage einer Korrelation zwischen dem Sendesignal (223) in dem Unterband und dem Signal (220) am entfernten Ende in dem Unterband;

- Bestimmung einer Schätzung des Echopegels des Signals (220) am entfernten Ende in dem Unterband durch Multiplizieren eines Pegels des Signals (220) am entfernten Ende in dem Unterband durch den Unterbandechopfadfilterkoeffizienten;

- Bestimmung einer Unterbanddämpfung für das Sendesignal (223) in dem Unterband durch Gewichtung des Unterbandechopfadfilterkoeffizienten mit dem Unterbandmaskierungsgewicht; und

- Dämpfung des Unterbands des Sendesignals (223) durch Subtrahieren des Pegels des Signals (220) am entfernten Ende in dem Unterband, multipliziert mit der Unterbanddämpfung von einem Pegel des Sendesignals (223) in dem Unterband.

Revendications

1. Un circuit de suppression d'écho (202) pour un dispositif électronique (100) comprenant un haut-parleur (107) et un microphone (105); dans lequel le circuit de suppression d'écho (202) est configuré pour atténuer une sous-bande d'un signal de transmission (223); dans lequel le signal de transmission (223) est soit capturé par le microphone (105) ou dérivé d'un signal capturé par le microphone (105); dans lequel le signal de transmission (313) comprend un écho d'un signal d'extrémité distante (220) rendu par le haut-parleur (107) et un signal d'extrémité proche (222); caractérisé en ce que le circuit de suppression d'écho (202) est configuré pour :

- déterminer un indicateur d'extrémité distante de sous-bande d'une activité vocale dans le signal d'extrémité distante (220) sur la base du signal d'extrémité distante (220) dans la sous-bande; dans lequel l'indicateur d'extrémité distante de sous-bande comporte un estimé d'un niveau de signal vocal compris entre le signal d'extrémité distante et/ou un indicateur de la vraisemblance que le signal d'extrémité distante comporte un signal vocal ;

- déterminer un indicateur d'extrémité proche de sous-bande d'une activité vocale du signal d'extrémité proche (222) sur la base du signal d'extrémité distante (220) dans la sous-bande et sur la base du signal de transmission (223) dans la sous-bande; dans lequel l'indicateur d'extrémité proche de sous-bande comporte un estimé d'un niveau d'un signal vocal compris à l'intérieur du signal d'extrémité proche et/ou un indicateur de la vraisemblance que le signal d'extrémité proche comporte un signal vocal ;

- déterminer un poids de masquage de sous-bande, basé sur l'indicateur d'extrémité distante de la sous-bande et basé sur l'indicateur d'extrémité proche de la sous-bande; dans lequel le poids de masquage de sous-bande indique un masquage de l'écho du signal d'extrémité distante (220) par le signal d'extrémité proche (222) dans la sous-bande;

- déterminer un coefficient de filtrage de chemin d'écho de sous-bande sur la base d'une corrélation entre le signal de transmission (223) dans la sous-bande et le signal d'extrémité distante (220) dans la sous-bande ;

- déterminer un estimé du niveau de l'écho du signal d'extrémité distante (220) dans la sous bande en multipliant un niveau du signal d'extrémité distante (220) dans la sous-bande par le coefficient de filtrage de chemin d'écho de sous-bande ;

- déterminer une atténuation de sous-bande pour le signal de transmission (223) dans la sous-bande par la pondération du coefficient de filtrage de chemin d'écho de sous-bande avec le poids de masquage de sous-bande ; et

- atténuer la sous-bande du signal de transmission (223) au moyen d'une soustraction du niveau du signal d'extrémité distance (220) dans la sous-bande multiplié par l'atténuation de sous-bande à partir d'un niveau du signal de transmission (223) dans la sous-bande.

2. Le circuit de suppression d'écho (202) de la revendication 1, dans lequel le circuit (202) est configuré pour:

- déterminer une pluralité d'indicateurs d'extrémité distante de sous-bande sur la base d'une pluralité correspondante de signaux d'extrémité distante de sous-bandes (220);

- déterminer une pluralité d'indicateurs d'extrémité proche de sous-bande sur la base d'une pluralité correspondante de sous-bandes du signal d'extrémité distante (220) et sur la base d'une pluralité correspondante de sous-bandes du signal de transmission (223);

- déterminer une pluralité de poids de masquage de sous-bande, sur la base de la pluralité d'indicateurs d'extrémité distante de sous-bande et sur la base de la pluralité d'indicateurs d'extrémité proche de sous-bande;

- déterminer, pour chaque sous-bande, un coefficient de filtrage de chemin d'écho de sous-bande sur la base d'une corrélation entre le signal de transmission dans ladite sous-bande et le signal d'extrémité distante dans ladite sous-bande ;

- déterminer une pluralité d'atténuations de sous-bande pour le signal de transmission (223) dans la pluralité de sous-bandes par la pondération, pour chaque sous-bande, du coefficient de filtrage de chemin d'écho de sous-bande de ladite sous-bande avec le poids de masquage de sous-bande de ladite sous-bandes ;

- atténuer la pluralité de sous-bandes du signal de transmission (223) en utilisant la pluralité correspondante d'atténuations de sous-bandes déterminées.

3. Le circuit de suppression d'écho (202) de la revendication 2, dans lequel le circuit (202) comprend

- une unité de transformation temps-fréquence configurée pour déterminer une pluralité de composantes de fréquence du signal d'extrémité distante (220) et du signal d'émission (223); et

- une unité fréquence-sous-bande configurée pour déterminer la pluralité de sous-bandes du signal d'extrémité distante (220) et du signal d'émission (223) parmi la pluralité de composantes de fréquence du signal d'extrémité distante (220) et du signal de transmission (223), respectivement.

4. Le circuit de suppression d'écho (202) de la revendication 3, dans lequel le circuit (202) est configuré pour

- déterminer une pluralité d'atténuations pour la pluralité de composantes de fréquence du signal de transmission (223) en interpolant la pluralité d'atténuations de sous-bande.

5. Le circuit de suppression d'écho (202) de l'une quelconque des revendications 2 à 4, dans lequel le circuit (202) comprend

- une unité de transformation temps-fréquence inverse configurée pour déterminer un signal de transmission supprimé par écho (223) parmi la pluralité de sous-bandes atténuées du signal d'émission (223).

6. Le circuit de suppression d'écho (202) de l'une quelconque des revendications précédentes, configuré en outre pour déterminer le poids de masquage de sous-bande sur la base d'un ratio de l'indicateur d'extrémité distante de sous-bande et sur du ratio d'extrémité proche de sous-bande.

7. Le circuit de suppression d'écho (202) de la revendication 6, configuré en outre pour déterminer le poids de masquage de sous-bande sur la base d'un facteur de correction adapté pour accorder une influence du masquage de l'écho du signal d'extrémité distante (220) par le signal d'extrémité proche (222) sur l'atténuation de sous-bande.

8. Le circuit de suppression d'écho (202) de l'une quelconque des revendications précédentes, dans lequel l'atténuation de sous-bande déterminée est une variante temporelle et / ou une variante de fréquence.

9. Le circuit de suppression d'écho (202) de l'une quelconque des revendications précédentes, configuré en outre pour déterminer le poids de masquage de sous-bande de telle sorte que l'atténuation appliquée à la sous-bande du signal de transmission (223) augmente, si le masquage de l'écho du signal d'extrémité distant (220) par le signal d'extrémité proche (222) dans la sous-bande diminue et inversement.

10. Un dispositif électronique (100) comprenant:

- un haut-parleur (107) adapté pour restituer un signal d'extrémité distant (220) reçu sur le dispositif électronique (100);

- un microphone (105) adapté pour capturer un signal de transmission (223); dans lequel le signal de transmission (223) capturé comprend un écho du signal d'extrémité distante rendue (220) et un signal d'extrémité proche (222);

- un circuit de suppression d'écho (202) de l'une quelconque des revendications 1 à 9, adapté pour atténuer une sous-bande d'un signal dérivé du signal capté (223), produisant ainsi un signal de transmission à suppression d'écho; et

- une unité de transmission configurée pour transmettre le signal d'émission à suppression d'écho à un émetteur du signal d'extrémité distante (220).

11. Le dispositif électronique (100) selon la revendication 10, comprenant en outre une unité d'annulation d'écho (201) adaptée pour

- déterminer une estimation de l'écho du signal d'extrémité distante rendue (220), sur la base du signal d'extrémité distante reçu (220). ; et

- soustraire l'estimation de l'écho du signal d'extrémité distante rendue (220) du signal capturé (223), produisant ainsi le signal dérivé du signal capturé (223).

12. Un procédé pour atténuer une sous-bande d'un signal de transmission (223); le procédé comprenant

- la capture du signal de transmission (223) en utilisant un microphone (105); dans lequel le signal de transmission (313) comprend un écho d'un signal d'extrémité distante (220) rendu par un haut-parleur (107) et un signal d'extrémité proche (222); ledit procédé étant caractérisé par :

- la détermination d'un indicateur d'extrémité distante de sous-bande d'une activité vocale dans le signal d'extrémité distante (220) sur la base du signal d'extrémité distante (220) dans la sous-bande;

dans lequel l'indicateur d'extrémité distante de sous-bande comporte un estimé d'un niveau d'un signal vocal compris à l'intérieur d'un signal d'extrémité distante et/.ou un indicateur de vraisemblance que le signal d'extrémité distante comporte un signal vocal ;

- la détermination d'un indicateur d'extrémité proche de sous-bande d'une activité vocale du signal d'extrémité proche (222) sur la base du signal d'extrémité distante (220) dans la sous-bande et sur la base du signal de transmission (223) dans la sous-bande ; dans lequel l'indicateur d'extrémité proche de sous-bande comporte un estimé d'un niveau d'un signal vocal compris à l'intérieur du signal d'extrémité proche et/ou un indicateur de vraisemblance que le signal d'extrémité proche comporte un signal vocal ;

- la détermination d'un poids de masquage de sous-bande, sur la base de l'indicateur d'extrémité distante de sous-bande et sur la base de l'indicateur d'extrémité proche de sous-bande; dans lequel le poids de masquage de sous-bande indique un masquage de l'écho du signal d'extrémité distante (220) par le signal d'extrémité proche (222) dans la sous-bande;

- la détermination d'un coefficient de filtrage de chemin d'écho de sous-bande sur la base d'une corrélation entre le signal de transmission (223) dans la sous-bande et le signal d'extrémité distante (220) dans la sous-bande ;

- la détermination d'un estimé du niveau de l'écho du signal d'extrémité distante (220) dans la sous-bande en multipliant un niveau du signal d'extrémité distante (220) dans la sous-bande par le coefficient de filtrage de chemin d'écho de sous - bande ;

- la détermination d'une atténuation de sous-bande pour le signal de transmission (223) dans la sous-bande en pondérant le coefficient de filtrage de chemin d'écho de sous-bande avec le poids de masquage de sous-bande ; et

- l'atténuation de la sous-bande du signal de transmission (223) en soustrayant le niveau du signal d'extrémité distante (220) dans la sous-bande multipliée par l'atténuation de sous-bande à partir d'un niveau du signal de transmission (223) dans la sous-bande.

Drawing